1. Field of the Invention
The present invention relates to online surveying to determine user preferences.
2. Description of the Prior Art
Prior art survey administration approaches and tools in-person, by telephone, hybrid methods (e.g.—using TV, radio, webcast to deliver items and using phones or SMS for submitting responses), self-administered (mail), self-administered (web apps via browsers).
Weaknesses in current approaches and tools include costs, speed, complexity, insufficient psychographic data, ineffective rewards, copyright/trademark infringement, lack of incorporation of social networking, and lack of optimization for mobile devices.
Prior art survey approaches are expensive because they require experts to design the survey and decide whom to target, and require a contract with an existing respondent pool or a recruitment effort to get sufficient respondents. Across current methods, administration costs still increase roughly in-line with the number of respondents required.
In-person administration has high costs per potential respondent contacted. Phone and hybrid methods tend to be cheaper per potential respondent contacted, but convert potential respondents into actual respondents at a lower rate. While self-administered methods tend to be substantially cheaper per potential respondent contacted than in-person, phone, or hybrid methods, they also have substantially lower conversion rates—forcing a much larger number of potential respondents to be contacted to achieve statistically significant response rates.
In-person recruitment is slow but has a relatively high response rate from potential respondents contacted when compared to other methods. By phone recruitment is also slow and requires many potential respondents to be contacted for each completed survey.
Self-administration—in all forms—has an even lower response rate from all potential respondents contacted than supervised methods, and is prone to unpredictable delays in completion for those respondents who do complete the surveys.
Survey design can be complex and difficult, requiring expert assistance. However, even when the survey design is very simple, administering surveys in a supervised context without substantially biasing responses requires domain-specific expertise. Methods of delivering self-administered surveys over broadcast media and or electronic networks can require technical expertise unrelated to survey design or analysis.
Demographic data is required for most surveys. While some surveys still collect the required demographic data from respondents as part of each survey administered, some newer survey administration approaches preserve collected demographic data for respondents. If the data is preserved for a respondent, they do not need to be asked for it again when they take additional surveys in the future. This can reduce the time required to take surveys, and increase response rate amongst repeat respondents.
Stored demographic data for past respondents can also improve the efficiency of recruitment efforts for future surveys with pre-emptive demographic targeting. For example, potential respondents whose demographics are already well represented in responses for a survey already in progress can be excluded from additional recruitment efforts.
With demographic data, there is a basic canon, including age, sex, ethnicity, location, education, profession and income. Once all this is collected, the net benefit of storing more or more detailed demographic information for a respondent drops off. These deeper demographic details may be important for a particular survey, but are unlikely to be broadly useful in recruitment for or analysis of future surveys.
Psychographic—or IAO (Interests, Activities and Opinions)—data is often collected as part of a survey. However, it is usually only collected in a tightly focused area specific to the survey administered. The respondent IAO data is used in the analysis of that survey, but is not stored in a way that associates the responses with the respondent for future reference. The IAO data can thus be considered episodic—which is to say it is only collected and used in the context of a single survey. This omission prevents past responses from being used as an aid in recruitment (psychographic targeting) and or analysis (correlation with past opinions without repeating the questions) for future surveys.
In contrast to demographic data, psychographic data tends to become more broadly useful—particularly for recruitment targeting—the more of it is available. The usual limitation on its collection is that longer surveys tend to have correspondingly lower response rates.
Prior art enticements and rewards tend to be non-dynamic and potentially biasing. Enticements must be revealed during recruitment—prior to survey administration—to influence whether a potential respondent will choose to participate in a survey. Enticements to participate in a survey tend to be generic (e.g. cash equivalents like AmEx gift cards) and uniform. When enticements are non-generic (i.e.—an item or service from an identifiable brand), they risk biasing the survey—both in terms of influencing who will agree to take the surveys and what opinions they might have regarding the brand of the gift or related brands. When enticements are non-uniform, the mechanics of the administration become more complex and the costs per response with current methods tend to rise because more experienced administrators are required for supervised administration methods and more complex automated systems are required for current hybrid and self-administered methods.
However, uniform enticements miss out on opportunities to adjust incentives based on potential or actual respondents matching targeted criteria. Easy-to-recruit demographics can be offered lower value rewards for participation, lowering overall administration costs. Hard-to-recruit demographics can be offered higher value rewards for participation, increasing response rates. Respondents with key demographic, psychographic or social networking characteristics (e.g.—having many friends, having a high propensity for sharing links, liking a particular organization, etc) can be offered “bonus” rewards either prior to survey administration as extra enticements to participate or after having completed a survey to improve their perception of a brand or organization. Conditional bonus rewards could also be offered as an incentive to take additional steps immediately upon survey completion. This allows surveys to be used as a way to camouflage what is essentially a targeted brand promotion message.
Survey construction can be constrained by trademark and or copyright limits on usage of brand-specific images or language. Usage of such content without permission in a printed survey can lead to objections from rights holders. Restrictions on usage of such content on web pages generally fall into three categories based on the method of inclusion and how much that method modifies the context of the content from that in which it was originally offered. Linking (i.e.—providing a hypertext link that can trigger the display of the external content in its original form) is generally permitted without prior permission. Framing (i.e.—the inclusion of external content within a web page such that standard browsers render both the page content and the external content together) is less clearly allowed, with one court finding that framing was a copyright infringement because the process resulted in an unauthorized modification of the linked site. (Futuredontics Inc. v. Applied Anagramic Inc., 45 U.S.P.Q. 2d 2005 (C.D. Cal. 1998). Inlining (i.e.—direct inclusion of external content in another web page mixed in with a given page) is usually considered more likely to be infringing than framing as the context has been even more clearly modified from its use on its site of origin. So, inclusion of external, rights-protected content in a web survey either via framing or inlining is likely to raise objections from the rights holders. As such, web-based survey administration tools which use web page UI elements (buttons, fields, etc) to collect responses are likely to raise objection when the survey items include trademarked or copyrighted images or phrases included either via framing or inlining. However, providing a survey—whether printed or in web form—that has links to external sites, where the rights-protected content can be viewed in its original form, should not require any prior permission and is unlikely to raise objections. While unlikely to be infringing, this makes taking the survey much more cumbersome, requiring the respondent to enter URLs or click back and forth between the external content and the page where their response is collected.
Taken together, the preceding limitations make constructing printed or web-based surveys collecting responses on icons, web sites or slogans from competing brands likely to be either potentially objectionable or unduly cumbersome.
Current survey administration systems do not use social networking and media tools as well as they could. Most current approaches do not use these tools at all. In addition to simplifying and automating the collection and or confirmation of demographic information, social graph data (personal and business connections) could be gathered as well as additional psychographic data (likes, dislikes, shared links). This additional information could be used to target potential respondents much more accurately, to identify tastemakers who are more likely to convince others to take surveys if they are so convinced, and to incentivize those more likely to be influential in recruitment more strongly. Current approaches do none of these things.
Current survey administration systems are not optimized for participation via mobile devices. While there have been methods described involving combinations of broadcast media (radio, TV) to send the questions out with mobile phones and or text messaging devices used by respondents to send their responses back, these require all respondents to be watching or listening to the questions at the same time—a requirement that limits the potential respondent pool dramatically.
Other described methods of mobile survey administration rely entirely on text messages, both to send out the questions and for respondents to return their responses. However, confining surveys to questions that can be delivered as text messages is substantially limiting, preventing questions involving images, for example.
While some web-based self-administered surveys can be taken via web-enabled mobile devices, the small form factor and limited user input methods common to these devices make taking these surveys substantially more cumbersome, driving response rates on these surveys for mobile users much lower.
While many mobile devices are capable of receiving “push notifications” (e.g.—email, text messages, alerts), these are not being used by current survey administration systems to make targeted users aware of newly available surveys they are likely to be interested in taking.
Collaborative filtering (CF) is the process of filtering for information or patterns using techniques involving collaboration among multiple agents, viewpoints, data sources, etc. Applications of collaborative filtering typically involve very large data sets. Collaborative filtering methods have been applied to many different kinds of data including sensing and monitoring data—such as in mineral exploration, environmental sensing over large areas or multiple sensors; financial data—such as financial service institutions that integrate many financial sources; or in electronic commerce and web 2.0 applications where the focus is on user data, etc.
Collaborative filtering is a method of making automatic predictions (filtering) about the interests of a user by collecting taste information from many users (collaborating). The underlying assumption of the CF approach is that those who agreed in the past tend to agree again in the future. For example, a collaborative filtering or recommendation system for television tastes could make predictions about which television show a user should like given a partial list of that user's tastes (likes or dislikes). Note that these predictions are specific to the user, but use information gleaned from many users. This differs from the simpler approach of giving an average (non-specific) score for each item of interest, for example based on its number of votes.
Factor Analysis (FA) is an approach for building a preference model that requires far fewer calculations to make a suggestion as compared to competing approaches. Factor analysis is a statistical method used to describe variability among observed variables in terms of a potentially lower number of unobserved variables called factors. In other words, it is possible, for example, that variations in three or four observed variables mainly reflect the variations in a single unobserved variable, or in a reduced number of unobserved variables. Factor analysis searches for such joint variations in response to unobserved latent variables. The observed variables are modeled as linear combinations of the potential factors, plus “error” terms. The information gained about the interdependencies between observed variables can be used later to reduce the set of variables in a dataset. Factor analysis originated in psychometrics, and is used in behavioral sciences, social sciences, marketing, product management, operations research, and other applied sciences that deal with large quantities of data.
Factor Analysis has been used with Collaborative Filtering to generate user preference models and vectors (See Collaborative Filtering with Privacy via Factor Analysis by John Canny. url: www.cs.berkeley.edu/˜jfc/′mender/sigir.pdf).